I keep score. Publicly.

Every forecast, graded after the event. The great calls, the near-misses, and the busts — all of it, on one page, forever.

Calls on the record

Verified hits

Honest busts

Times the atmosphere won anyway

Graded forecasts, most recent first
Date	Event	The call	What happened	Grade
2026-01-16	The Blizzard of the Long Night	24–30" with a sharp northern gradient	27.4", gradient 6 miles north	HIT
2026-04-04	The April Dryline Outbreak	Discrete supercells then a line after dark	Both happened, 70 miles east	CLOSE
2026-06-12	Southern Plains severe window	Two tornadic supercells by 7pm	Cap never broke. Nothing.	BUST
2026-07-20	Single-band lake effect	2–3"/hr for 11 hours, three towns	11 hours, same three towns	HIT
2026-02-08	Ice storm, southern tier	0.4–0.6" accretion	0.55" accretion	HIT
2026-03-02	Coastal low track	Rain, not snow, inside the beltway	Snow for two hours, then rain	CLOSE
2026-05-19	Hail event, high plains	Golf ball to baseball, isolated	Baseball, and more of them than called	CLOSE
2025-11-27	Holiday travel forecast	Dry Wednesday, wet Thursday	Dry Wednesday, wet Thursday	HIT
2025-12-15	First measurable snow	2" Sunday night	Trace. It trended east.	BUST
2026-01-02	Arctic outbreak duration	Six days below zero overnight	Seven days	HIT
2026-04-22	Flash flood risk, urban corridor	3–5" in six hours	4.1" in five hours	HIT
2026-06-30	Derecho potential	Wind risk stays north of the river	It did not stay north of the river	BUST

Because “trust me” isn't a forecast

Anyone can post a confident take before a storm. The record is what separates a forecaster from a highlight reel. The atmosphere grades on a brutal curve, and I publish the report card anyway.

How grading works

Each call is scored against what actually happened — timing, placement, magnitude. No partial credit for vibes. No quietly deleting the misses. The methodology is fixed before the event, so I can't move the goalposts after.